Agents need constitutional authority before they gain operational power.
A permission layer defining observer, draft, review, operator, technician, constitutional, recovery, and adversarial test agent scopes.
Every TheoB pathway can move through Past, Present, and Future without losing context.
Read current signals, conditions, and live context.
Voice ready
Helpful agents become dangerous when authority is fuzzy.
TheoB agents should know what they can observe, draft, review, prepare, execute, block, recover, and stress-test — without self-expanding, exposing secrets, or touching production-critical systems without approval.
May watch signals, logs, routes, and health without changing infrastructure.
May prepare patches, plans, audits, and recommendations without deployment authority.
May flag risks, compare changes, score confidence, and request human review.
May execute approved low-risk workflows within bounded scope.
May prepare deployment, runtime, key, and provider actions but requires approval for critical changes.
May enforce boundaries, block unsafe actions, and escalate protected-value violations.
May prepare rollback, fallback, restore, and incident recovery plans.
May simulate stress, attack patterns, manipulation, overload, and bad actor behavior in sandbox only.
Agent power must be scoped before it is useful.
Agents cannot expand their own authority.
Critical infrastructure actions require founder or approved human review.
Secrets may be referenced by purpose, never exposed raw.
Agent permissions must be scoped by domain, risk, and reversibility.
Sandbox stress testing must never spill into production systems.
Every agent action should produce audit memory.